High-entropy alloys serve as the prominent candidates for applications where superior mechanical strength and corrosion resistance of the used materials is a must. It is believed that such unique properties of these compounds are due to high configurational entropy, provided by a large number of constituent chemical elements. Owing to this, exploring the chemical space for possible compounds using electronic structure methods such as DFT is a rather intractable approach, as complexity of the latter dramatically increases with the number of particles and chemical elements within a system.
In this work, we propose a machine-learning method for modeling high-entropy alloys. Using high-entropy carbide HfTaTiNbZrC5 as an example, we show that machine-learning interatomic potential coupled with Monte Carlo simulations is capable of modeling phase transitions within crystal structure at a much lower computational cost. Moreover, obtained results are in good agreement with experimental findings, which makes this method prominent for such types of materials.